Vaginal microbiota varies by geographical location in South African women

Full text

 hp://www.satnt.ac.za 1 OpenAccess
 Pageiofii Inhoudsopgawe Page1of9 OorspronklikeNavorsing
South African Journal for Science and Technology
ISSN: (Online) 2222-4173, (Print) 0254-3486
Vaginal microbiota varies by geographical locaon
in South African women
Authors:
KaeLennard1,Smritee
Dabee2,ShaunLBarnabas2,
EnockHavyarimana2,
AnnaBlakney3,Shameem
ZJaumdally,GerritBotha1,
NonhlanhlaNMkhize7,
Linda-GailBekker4,
DavidALewis5,6,7,Glenda
Gray8,9,NicolaMulder1,
Jo-AnnSPassmore2,10,
HeatherBJaspan2,11*
Aliaons:
1InstuteofInfecous
DiseaseandMolecular
MedicineandComputa-
onalBiologyDivision,
DepartmentofIntegrave
BiomedicalSciences,
UniversityofCapeTown,
SouthAfrica
2InstuteofInfecous
DiseaseandMolecular
MedicineandDepartment
ofPathology,Universityof
CapeTown,SouthAfrica
3DepartmentofBioengi-
neering,Universityof
Washington,UnitedStates
4InstuteofInfecous
DiseaseandMolecular
Medicine,DesmondTutu
HIVCentre,Universityof
CapeTown,SouthAfrica;
5WesternSydneySexual
HealthCentre,Australia
6MarieBashirInstutefor
InfecousDiseasesand
BiosecurityandSydney
MedicalSchool-Westmead,
UniversityofSydney,
Australia
7NaonalInstuteforCom-
municableDiseases,South
Africa
8PerinatalHIVResearch
Unit,UniversityoftheWit-
watersrand,SouthAfrica
9SouthAfricanMedical
ResearchCouncil,Cape
Town,SouthAfrica
10NaonalHealthLabora-
toryService,SouthAfrica
11SealeChildren’sResearch
Instute,Department
ofPediatricsandGlobal
Health,Universityof
Washington,USA
Corresponding author:
DrHeatherJaspan
hbjaspan@gmail.com
Dates:
Received:17/09/2018
Accepted:12/04/2019
Published:
Women of African descent are more likely to have bacterial vaginosis than women of
other ethnicities. Both diversity and likely specic taxa in these microbial communities are
important to sexual and reproductive health, such as HIV risk. However, whether the specic
taxa also vary by geographical location and/or ethnicity requires further investigation.
Here, we compare the vaginal microbiota of 16–22-year-old black, HIV-negative South
African women from two geographically disparate but low-income high population density
communities, one in Cape Town (CPT) and one in Johannesburg (JHB). Vaginal microbiota
composition was assessed by 16S rRNA gene amplicon sequencing of lateral vaginal wall swabs.
Geographical location was signicantly associated with vaginal microbiota composition
by permutational analysis of variance (PERMANOVA) (p=0.02), as were body mass index
BMI (p=0.015) and human papilloma virus (HPV) risk type (p=0.005), while the presence
of one or more sexually transmitted infections (STIs) (p=0.053) and hormonal contraceptive
(HC) usage (p=0.4) were not. Geographical location remained a signicant determinant of
microbiota composition independent of BMI, STI status and HPV-risk. Together, geographical
location, BMI and HPV-risk explained 10% of the variance in microbiota composition with a
large proportion of the variance remaining unexplained. Several taxa differed signicantly
between geographical location – some by frequency and others by relative abundance.
Our results therefore suggest that HIV prophylactic approaches targeting the vaginal
microbiota should be geographically tailored.
Geograese ligging beïnvloed vaginale mikrobiese proele in Suid Afrikaanse vroue:
Vroue van Afrika-afkoms is meer vatbaar vir bakteriële vaginose (BV) in vergelyking met
Europese vroue. Beide mikrobiese diversiteit (soos met BV) sowel as spesieke bakteriële
taksa speel ‘n rol in seksuele en reproduktiewe gesondheid insluitende MIV vatbaarheid.
Die moontlike rol van geograese ligging en etnisiteit op die verhouding tussen
mikrobiese samestelling en seskuele en reproduktiewe gesondheid bly egter onbekend.
In hierdie studie vergelyk ons dus die vaginale mikrobiota van 16–22-jarige swart, HIV-
negatiewe Suid Afrikaanse vroue van twee geograes-uiteenlopende liggings, beide lae-
inkomste,hoë bevolkingsdigtheidsgemeenskappe, een in Kaapstad, en een in Johannesburg.
Vaginale mikrobiese proele is bepaal met behulp van 16S rRNS volgordebepaling van
laterale muur deppers.
Ons pas permutasie variansieanalise (PERMANOVA) toe en vind statisties betekenisvolle
assosiasies tussen vaginale mikrobiese samestelling en geograese ligging (p=0.02), asook
met liggaamsmassa-indeks (LMI) (p=0.015) en menslike papilloomvirus (MPV) risikotipe
(p=0.005), maar nie met die voorkoms van een of meer seksueel-oordraagbare infeksies
(SOI’s) (p=0.053) of met hormonale kontrasepsie verbruik nie.(p=0.4)
Geograese ligging was ‘n statisties betekenisvolle determinant van mikrobiese
samestelling, ongeag verskille in LMI, SOI status en MPV-risiko tipes tussen Kaapstad
en Johannesburg vroue. Geograese ligging, LMI en MPV-risiko verduidelik gesamentlik
10% van die variansie in mikrobiese samestelling, met ‘n groot persentasie van onbekende
oorsprong. Verskeie taksa het statisties betekenisvol verskil in terme van frekwensie of
relatiewe vlakke van voorkoms tussen die geograese liggings.
Ons resultate stel voor dat MIV prolaktiese metodes wat die vaginale mikrobiota teiken
die effek van geograese ligging in ag moet neem.
How to cite this arcle:
KaeLennard,SmriteeDabee,ShaunLBarnabas,EnockHavyarimana,AnnaBlakney,ShameemZJaumdally,
GerritBotha,NonhlanhlaNMkhize,Linda-GailBekker,DavidALewis,GlendaGray,NicolaMulder,Jo-AnnS
Passmore,HeatherBJaspan,VaginalmicrobiotavariesbygeographicallocaoninSouthAfricanwomen,Suid-
Afrikaanse Tydskrif vir Natuurwetenskap en Tegnologie 38(1) (2019)
’nAfrikaansevertalingvandiemanuskripisaanlynbeskikbaarbyhp://www.satnt.ac.za/index.php/satnt/
arcle/view/685
Copyright:
©2019.Authors.
Licensee:Die Suid-Afrikaanse Akademie vir Wetenskap en Kuns.ThisworkislicensedundertheCreave
CommonsAbuonLicense.

 hp://www.satnt.ac.za 2 OpenAccess
 Page2of9 OorspronklikeNavorsing
Johannesburg (JHB). Approval was obtained for the study
from the Research Ethics Committees of the Universities of
Cape Town and Witwatersrand. All participants ≥ 18 years
provided informed consent, while assent and parental
consent were obtained for participants ≤ 18 years. Young
women were enrolled if they were HIV-negative, in general
good health, not pregnant or menstruating at the time
of sampling, and if they had not had unprotected sex or
douched in the last 48 hours. Additional exclusion criteria
were use of antibiotics in the prior two weeks. Study visits
were scheduled two weeks after injection for participants
on injectable progestin contraceptives, or otherwise during
the luteal phase of their menstrual cycles (between day
14–28) if they were not using any HCs or if they were using
oral HCs. Before specimen collection, the following were
performed: HIV pre-test and risk-reduction counselling, an
HIV rapid test (Alere Determine™ HIV-1/2 Ag/Ab Combo,
Alere, Waltham, MA), a pregnancy test (U-test Pregnancy
strip, Humor Diagnostica, Pretoria, South Africa) and a
general physical examination. Cervicovaginal uid via
disposable menstrual cup (Softcup®), one vulvovaginal
swab for STI testing and one lateral vaginal wall swab for
Nugent scoring and microbiome analysis were collected.
STI and BV tesng
Vulvo-vaginal swabs were assayed for nucleic acid of the
following STIs by multiplex PCR: Chlamydia trachomatis,
Neisseria gonorrhoeae, Trichomonas vaginalis, Mycoplasma
genitalium, HSV-1 and -2, Haemophilus ducreyi, Treponema
pallidum and lymphogranuloma venereum as previously
described (Lewis et al. 2012). Endo-cervical swabs were
collected for HPV detection and genotyping by Roche
Linear Array (Mbulawa et al. 2018). The following HPV
types were considered high-risk HPV: 16, 18, 31, 33, 35, 39,
45, 51, 52, 56, 58, 59, 66, 68 (Jacobs et al. 1997). For analyses
referring to STI (any), women considered positive had at
least one of the STIs tested for in this study, excluding HPV.
Lateral wall/posterior fornix swabs were collected for
Nugent scoring to classify samples as BV negative (Nugent
0–3), intermediate (Nugent 4–6) or positive (Nugent 7–10);
and vaginal pH was measured using colour-xed indicator
strips (Macherey-Nagel, Düren, Germany).
16S sequencing and analysis
Swabs were thawed, treated with a cocktail of mutanolysin
(25kU/ml, Sigma Aldrich), lysozyme (450kU/ml, Sigma
Aldrich), and lysostaphin (4kU, Sigma Aldrich), then
mechanically disrupted with a bead-beater. DNA was
extracted using the MoBio PowerSoil DNA extraction kit
(MoBio, Carlsbad, CA). The V4 region of the 16S rRNA
gene was amplied using universal primers that were
modied to encode the Illumina MiSeq sequencing primer
sequence at the 5’ end (46): 515F (TCG TCG GCA GCG TCA
GAT GTG TAT AAG AGA CAG NNN NNG TGC CAG
CMG CCG CGG TAA) and 806R (GTC TCG TGG GCT
CGG AGA TGT GTA TAA GAG ACA GNN NNN GGA
Introducon
Vaginal microbiota proles vary by ethnicity (Srinivasan
et al. 2012; Buvé et al. 2014; Ravel et al. 2010). Women
of African descent less commonly have Lactobacillus-
dominant vaginal microbiota compared with Caucasian
women (Ravel et al. 2010; Anahtar et al. 2015b; Lennard et
al. 2017). This nding appears to be generalisable to African
American and Hispanic women from North America who
frequently have decreased relative abundance of
(Anahtar et al. 2015a; Fettweis et al. 2014; Zhou et al.
2007). It is less clear to what extent geographical location
affects vaginal microbiota composition among women of
the same ethnicity. Bacterial vaginosis (BV) rates vary by
ethnicity and geographical location (with potentially large
variation in the proportion of BV among different African
countries) (Kenyon, Colebunders, and Crucitti 2013). Yet,
detailed description of vaginal microbiota composition by
geographical location is currently lacking.
It has long been recognised that bacterial vaginosis (a
vaginal dysbiosis) is associated with adverse sexual
and reproductive health outcomes, including sexually
transmitted infections (Wiesenfeld et al. 2003; Gallo et
al. 2012; Balkus et al. 2014) and adverse birth outcomes
(Leitich and Kiss 2007; Holst, Goffeng, and Andersch
1994; Nelson et al. 2015). Recently, with the advent of next
generation sequencing, specic taxa have been implicated
in these outcomes – such as preterm births (Freitas et al.
2018; Tabatabaei et al. 2018; Vinturache et al. 2016) and
HIV risk (McClelland et al. 2018). In a study conducted
on ve separate cohorts from Kenya, Uganda, South
Africa, Tanzania, Botswana and Zambia, McClelland et
al. identied taxa that were associated with increased
odds of HIV acquisition across all cohorts considered,
some of which were signicantly so (Parvimonas species
type 1 and 2, Gemella asaccharolytica, Mycoplasma hominis)
(McClelland et al. 2018). The question as to whether we
can dene a robust microbiota signature of HIV risk that is
generalisable across geographical locations/ethnicities, or
whether location-specic taxa should be identied for HIV
risk assessments remains.
Here, we compare the vaginal microbiota of 16–22-year-old
black HIV-negative South African women from two low-
income high population density communities; one in Cape
Town (CPT) and the other Johannesburg (JHB).
Materials and Methods
Parcipant selecon and sample collecon
The cohorts have been described previously in detail
(Barnabas et al. 2017; Lennard et al. 2017). Briey, 298 black,
16–22-year-old HIV-negative South African women were
recruited as part of the Women’s Initiative in Sexual Health
(WISH) study (Barnabas et al. 2017) from low-income, high
population density communities in Cape Town (CPT) and

 hp://www.satnt.ac.za 3 OpenAccess
 Page3of9 OorspronklikeNavorsing
CTA CHV GGG TWT CTA AT), where NNNNN indicates
ve randomly incorporated nucleotides for increased
complexity (Pearce, Hilt, and Rosenfeld 2014). The 5’ end is
the Illumina Nextera adapter, and the sequences following
the Ns are complementary to the V4 rRNA gene region.
Pooled samples were puried with AMPure XP beads
(Beckman Coulter, Brea, CA, USA) and quantied by
using the PicoGreen double-stranded DNA (dsDNA)
assay (Invitrogen, Carlsbad, CA, USA). Dual indices and
Illumina sequencing adapters were attached using the
Nextera XT DNA Prep kit (Illumina). Samples were again
puried by using AMPure XP beads, quantied by using a
Qubit uorometer (Invitrogen), and pooled for sequencing.
Puried libraries consisting of 96 pooled samples were
paired-end sequenced on an Illumina MiSeq platform (300-
bp paired-end reads with V3 chemistry).
Following demultiplexing, raw reads were preprocessed
as follows: forward and reverse reads were merged
using usearch7 (Edgar 2010), allowing a maximum of
three mismatches; merged reads were quality ltered
using usearch7 (reads with E scores larger than 0.1 were
discarded); primer sequences were removed using a
custom python script; and merged, ltered reads were
truncated at 250bp. Next, sequences were de-replicated
whilst recording the level of replication for each sequence
using usearch7. De-replicated sequences were sorted by
abundance (highest to lowest) and clustered de novo into
operational taxonomic units (OTUs) at 97% similarity
using usearch7. Chimeric sequences were detected (against
the Gold database) using UCHIME (Edgar et al. 2011)
and removed. Individual sequences were assigned to
the specic identiers using a 97% similarity threshold.
Taxonomic assignment was performed in QIIME 1.8.0
(Caporaso et al. 2010) using the RDP classier (using the
default condence level of 0.5) against the GreenGenes 13.8
reference taxonomy for 97% identity. To increase species-
level resolution, we constructed a custom taxonomic
database appropriate for V4 region 16S rRNA gene
amplicon sequencing based on the custom vaginal 16S
rRNA gene reference database created by Fettweis et al.
(Fettweis et al. 2012). This database was updated for the
V4 region and used to increase species-level resolution
as previously described (Lennard et al. 2017). OTUs that
mapped to more than one species (with the same identity
score) were annotated as follows: if an OTU mapped to two
or three species, the OTU would be named Genus speciesA_
speciesB or Genus speciesA_speciesB_speciesC, respectively,
and if an OTU mapped to more than three species but one
species was clearly associated with vaginal microbiota
(based on prior knowledge), the OTU was named Genus
species_cluster, where “species” was selected based on the
majority of hits; e.g., L. reuteri_cluster indicates the case
where the majority of hits were for L. reuteri but there were
several other species with equal identity scores present.
Samples with ≥ 5000 reads were selected for downstream
analyses. The OTU table was standardised (i.e. transformed
to relative abundance and multiplied by the median sample
read depth), and ltered so that each OTU had to have at
least 10 counts in at least 2% of samples or have a relative
abundance of at least 0.001%.
Stascal analyses
All downstream statistical analyses were performed in R,
using the packages phyloseq (McMurdie and Holmes 2013)
for beta diversity analyses, metagenomeSeq (Paulson et al.
2013) for differential abundance testing, vegan (Oksanen
et al. 2016) for ordinations and redundancy analysis, and
NMF (Gaujoux 2014) for annotated heat maps.
Permutational multivariate analysis of variance (PER-
MANOVA) was performed using the adonis and adonis2
functions from the R package vegan (Oksanen et al. 2016);
for the adonis function the order of predictor variables
matter, while the order of terms do not affect results
in the adonis2 function. Because we did not wish to
make assumptions regarding the relative importance of
predictor variables, adonis2 was used to obtain p-values
for individual variables, while adonis() was used to
obtain adjusted R2 values (which are not available when
using adonis2). The assumption for PERMANOVA of
homogeneity of variance between groups was assessed
using the betadisper() function from the R package vegan
(Oksanen et al. 2016). This assumption was met when
using Bray-Curtis as distance metric, but not when using
UniFrac distance or weighted UniFrac distance; hence we
used Bray-Curtis distance. In the nal model ethnicity
(for which 24 participants had missing information) was
excluded as ethnicity was not signicant when performing
PERMANOVA on the subset of participants for whom we
did have ethnicity information.
Distance-based redundancy analysis (db-RDA) was
performed on Bray-Curtis dissimilarity matrix using
the dbrda() function from the R package vegan and the
ordination was constrained on geographical location, STI
(any), BMI and HPV risk (variables that were not signicant
by PERMANOVA were excluded from the nal model
used for visual presentation in Figure 1).
Differences in microbial composition between groups of
interest were assessed using the R package metagenome-
Seq’s MRfulltable function with a custom lter to determine
signicance: merged taxa were deemed signicantly
different if they exhibited a fold change (beta coefcient) of
≥ 1.5, had an adjusted p-value of ≤ 0.01 and if at least one
of the two groups being compared had ≥ 20% of samples
with the given taxon OR the Fisher’s exact test result
was signicant (after multiple testing correction). OTUs
were rst merged at the lowest available taxonomic level

 hp://www.satnt.ac.za 4 OpenAccess
 Page4of9 OorspronklikeNavorsing
(i.e. for OTUs with Lactobacillus as the lowest available
taxonomic annotation counts were summed, while OTUs
with additional species-level annotation e.g. L. iners were
summed at species-level instead). Composite barplots
(Figure 2) were also created based on this merged table.
The most abundant taxa were selected as follows: For each
sample the most abundant taxa were determined (based
on standardised, merged taxon counts), after ranking taxa
for each sample by read counts (high to low) and selecting
those taxa that cumulatively made up the rst 50% of reads
for that sample. This resulted in a list of 28 unique taxa
across all samples, which was then limited to the subset that
had been classied as ‘abundant’ in at least two samples,
reducing the number of abundant taxa to 12 (Figure 2).
Random forests analyses were conducted on merged taxa to
determine which taxa best predicting geographical location
using the R packages randomForest (Liaw and Wiener
2002) and ROCR (Sing et al. 2005) for ROC analysis. The full
dataset in question was used to train random forests models,
i.e. the data were not divided into training and test sets.
Results
Microbiota proling was performed by 16S rRNA gene
amplicon sequencing for 102 women from CPT and 79
women from JHB (Table 1). The two groups were well
matched in terms of age (median 18 years for both locations).
Hormonal contraceptive usage differed signicantly with
100% of CPT women compared with 41% of JHB women
on some form of hormonal contraceptive, likely due to
differences in recruitment approaches between the two sites
(CPT participants were recruited through a family planning
clinic while JHB participants were recruited from a broader
population). CPT women had higher BV prevalence (55 vs.
35%), STI prevalence (59 vs. 24%), BMI (25.4 vs. 22.5), had
higher levels of genital inammation and were of more
homogenous ethnicity than JHB women (Table 1).
To identify factors inuencing vaginal microbiota pro-
les, permutational multivariate analysis of variance (PER-
MANOVA) was performed. Factors considered included
ethnicity, age, hormonal contraceptive usage (yes/no),
the presence of any one or more STI excluding HPV (yes/
no), HPV-risk (high/low/negative), geographical location
and BMI. Ethnicity was not included in the nal model
since there were 24 participants for whom we did not have
ethnicity information and ethnicity was not a signicant
factor when performing PERMANOVA on the subset for
whom we did have ethnicity information. Age was also
not included in the nal model as there was no signicant
different in age between JHB and CPT (Table 1). Factors
signicantly associated with vaginal microbiota composition
were geographical location (p=0.02), BMI (p=0.015), and
TABLE 1: Parcipantsummarybygeographicallocaon
Feature Cape Town (N=102) Johannesburg (N=79) P value
Medianage,years 18 18 0.6
BVprevalence,n(%)   0.008
BV posive 56(55) 28(35)
BV intermediate 7(7) 15(19)
BVnegave 39(38) 36(46)
Nugentscore(median) 8 4 0.01
STI(any),n(%) 60(59) 19(24) 2.7e-6
C. trachomas 45(44) 13(17) 1e-4
N. gonorrhoeae 14(14) 4(5) 0.08
T. vaginalis 6(6) 3(4) 0.7
M. genitalium 4(4) 2(3) 0.7
HSV-2 (DNA) 6(6) 1(1) 0.1
HPVrisk,n(%)   0.3
High 42(41) 33(42)
Low 29(28) 15(19)
Negave 31(30) 31(39)
Hormonalcontracepves&,n(%)   < 2.2e-16
DMPA  19(19) 9(12)
Implanon 8(8) 0(0)
Nur isterate 70(69) 12(15)
OCP 4(4) 6(8)
Male condom 0(0) 36(46)
Nuvaring 1(1) 0(0)
Injectable (type not specied) 0(0) 5(6)
None 0(0) 10(13)
Usinghormonalcontracepves,n(%)   < 2.2e-16
Any hormonal contracepve  102(100) 32(41)
Condoms/none  0(0) 46(59)
Ethnicity#,n(%)
Xhosa 94(99) 17(28) < 2.2e-16
OtherQ 1(1) 44(72)
BMI (median) 25.3 22.5 0.04
Y=Yes;N=No;&onewomanhadincompleteHCdata;#25womenhadincompleteethnicitydata

 hp://www.satnt.ac.za 5 OpenAccess
 Page5of9 OorspronklikeNavorsing
HPV-risk (p=0.005), while STI (p=0.053) and HC use (p=0.4)
were not. Together these factors explained ~10% of the
variation in microbiota composition with the remaining 90%
of unknown origin. Given the large discrepancy in HC use
between geographical locations (Table 1) PERMANOVA
was redone on the subset of women who used HC, excluding
those who used condoms only or no form of contraceptive.
Again location (p=0.02), BMI (p=0.03) and HPV-risk (p=0.05)
were signicant while STI use was not (p=0.7).
Distance-based redundancy analysis (db-RDA) was
performed on the Bray-Curtis dissimilarity matrix, which
conrmed the PERMANOVA results (Figure 1); db-RDA is
a constrained principal coordinates analysis, which allows
the use of non-Euclidean dissimilarity indices such as Bray-
Curtis, therefore more suited to 16S rRNA gene microbiome
data. Factors included in the db-RDA ordination included
geographical location, STI other than HPV (yes/no),
HPV-risk and BMI (i.e. factors that were signicant by
PERMANOVA, with STIs p=0.053). To further conrm
the signicance of geographical location on microbiota
composition, factors that vary signicantly by location
(STI(yes/no), HPV-risk and BMI) were partialed out in the
db-RDA model, yet location remained signicant (p=0.02).
The most abundant taxa are summarised by geographical
location in Figure 2.
To determine which taxa signicantly differed between CPT
and JHB, differential abundance analysis was performed
using the R package metagenomeSeq. The analysis
was performed on taxa merged at the lowest available
FIGURE 1: Distance-basedredundancyanalysis(dbRDA) ontheBray-Cursdissimilaritymatrixby geographicallocaon,STI(yes/no),HPV-risk(high/medium/low)
andBMI

 hp://www.satnt.ac.za 6 OpenAccess
 Page6of9 OorspronklikeNavorsing
taxonomic annotation (see Methods for details). Eighteen
taxa differed signicantly in terms of frequency and/or
relative abundance between JHB and CPT (Figure 3).
Taxa that were signicantly higher in frequency (i.e.
proportion of positive samples) in CPT compared to JHB
included Bidobacterium, Prevotella pallens, Pseudomonas,
Elizabethkingia meningoseptica, Brevundimonas, Myco
plas mataceae and Chryseobacterium whereas Lacto bacillus
coleohominis, Lactobacillus reuteri_cluster, Morganella morganii
and Varibaculum cambriense were more common in JHB
women. Taxa that were present at similar frequencies
between the CPT and JHB but varied in terms of relative
abundance were Leptotrichiaceae, Sneathia sanguinegens, P.
amnii, Prevotella and BVAB3 (Mageeibacillus indolicus), all of
which had higher relative abundance in samples from CPT.
Random forest analysis identied M. morganii and V.
cambriense as the highest ranked taxa to distinguish samples
from JHB vs. CPT (training AUC=0.95, PPV=0.91, NPV=0.89).
Discussion
Vaginal microbiota proles are known to vary by ethnicity
and geographical location. Here we demonstrate differences
in the relative abundance and frequency of colonisation of
specic vaginal microbiota in African women from CPT
and JHB, of similar ages and socioeconomic backgrounds.
These differences could not fully be explained by factors
that differed by geographical location, including hormonal
contraceptive usage, ethnicity, BMI, HPV-risk or the
occurrence of STIs. Together, geographical location, BMI
and HPV-risk explained 10% of the variance in microbiota
composition with a large proportion of the variance
remaining unexplained.
McClelland et al. found in ve different African cohorts,
that the concentration of certain taxa were associated with
later HIV seroconversion. In sensitivity analyses using
frequency, however, certain of these taxa were clearly
of more importance in specic cohorts. For example, the
FIGURE 2: CompositebarplothighlighngthemostabundanttaxainA)CPTand B)JHB.Taxaincludedinthelegendwereselected fromthosetaxathatmadeup
therst50% ofeachsample whenrankedbyabundance.Parcipants areorderedbasedon theirmostdominanttaxon,matching theorderofthe gurelegend
(e.g.parcipantsforwhomMegasphaerawasthemostabundanttaxonarelistedrst);other:thesummedabundanceofalltaxanotincludedinthegurelegend.

 hp://www.satnt.ac.za 7 OpenAccess
 Page7of9 OorspronklikeNavorsing
presence of detectable Mycoplasma hominis played a role
in HIV risk in Kenyan female sex workers but not in
serodiscordant couples from Uganda and South Africa,
where Gemella and Parvimonas were more important players
in the latter cohort (McClelland et al. 2018). In a study of
women from KwaZulu Natal, Williams et al found that
the relative abundance of Prevotella bivia to be the taxon
most predictive of later HIV seroconversion (Williams,
AIDS Conference 2016). Finally, Gossman et al, also in a
cohort from KZN but younger than the CAPRISA cohort,
found that relative abundances of P. melaninogenica and
Veillionella montpellierensis were the taxa most predictive of
later HIV seroconversion (Gosmann et al. 2017).
Several studies of vaginal microbiota and preterm birth
have found Lactobacillus-dominant vaginal microbiota
to be protective, however, no taxa have consistently been
associated with increased risk of this outcome (Dingens
2016, Romero 2014). Freitas et al. found the concentration of
Mollicutes to be a potential risk factor (Freitas et al. 2018).
Although Gardnerella and Ureaplasma relative abundance
were predictive of preterm birth in a predominantly
Caucasian cohort from California, Callahan et al. were
unable to replicate these ndings in a predominantly African
American cohort from Alabama (Callahan et al. 2017).
In summary, although there may be a subset of taxa
consistently associated with adverse sexual and repro-
ductive outcomes across a range of geographical loca-
tions, several clinically relevant taxa may be missed if
geographical context is ignored. It remains unclear what
might be driving these geographical differences in vaginal
microbiota composition – environmental/community micro-
biota composition during early-life establishment of the
microbiome likely plays an important role. Independent of
the origin of these differences, our results strongly argue
for geographically tailored microbiome-based diagnostics
and therapeutics, even within the same country.
FIGURE 3: Taxa(mergedatlowestavailabletaxonomicannotaon)thatdieredsignicantlyintermsofrelaveabundanceand/orfrequencybetweenCPTandJHB.
Columnshavebeenmanuallysortedbylocaon,buthierarchicalclusteringwasperformedwithineachlocaonsubset.Resultswerelteredasfollows:FDR≤0.01,
betacoecient≥1.5andeachtaxonhadtobepresentinatleast20%ofsamplesfromeitherorbothlocaons.

 hp://www.satnt.ac.za 8 OpenAccess
 Page8of9 OorspronklikeNavorsing
Acknowledgements
This study was supported by grants from the European
and Developing Countries Clinical Trials Partnership
(EDCTP) Strategic Primer grant [SP•2011•41304•038] and
the South African Department of Science and Technology
[DST/CON 0260/2012]. KL was supported by the National
Research Foundation and the Suid-Afrikaanse Akademie
vir Wetenskap en Kuns. HBJ was supported in part by
K08HD069201. SLB was supported by the HIV Vaccine
Trials Network SHAPe Program, the Fogarty Foundation
and the South African Medical Research Council (MRC).
SD was supported by the National Research Foundation
of South Africa. The DTHF also recognises the support
from ViiV health care in their YouthShield program. PHRU
was supported through funding from the South African
Medical Research Council. We thank the WISH Study
Teams, particularly Ms Pinky Ngobo, Sr Nozipho Hadebe,
Sr Janine Nixon, and all the young women who kindly
participated in the study. We thank Prof Lynn Morris,
David Lewis, Venessa Maseko and Raveshni Durgiah
from the National Institute for Communicable Diseases
for their help with sample processing. Computations were
performed using facilities provided by the University of
Cape Town’s ICTS High Performance Computing team:
http://hpc.uct.ac.za
References
Anahtar, M. N., Byrne, E.H., Doherty, K. E., Bowman, B. A., Yamamoto, H. S.,
Soumillon, M., Padavaan, N., et al. 2015a. CervicovaginalBacteria Are a
Major Modulator of Host Inammatory Responses in the Female Genital
Tract. Immunity 42 (5). Elsevier Inc.:965-76. hps://doi.org/10.1016/j.
immuni.2015.04.019.
Anahtar, M.N., Byrne, E. H., Doherty, K. E., Bowman, B. A., Yamamoto, H. S.,
Soumillon, M., Padavaan, N., et al. 2015b. CervicovaginalBacteria Are a
Major Modulator of Host Inammatory Responses in the Female Genital
Tract. Immunity 42 (5). Elsevier Inc.:965-76. hps://doi.org/10.1016/j.
immuni.2015.04.019.
Balkus, J.E., Richardson, B. A., Rabe, L. K., Taha, T.E., Mgodi, N., Kasaro, M. P.,
Ramjee,G., Homan,I.F.,and Karim,S. S.A. 2014.Bacterial Vaginosisand
theRiskofTrichomonasVaginalisAcquisionamongHIV-1-NegaveWomen.
Sexually Transmied Diseases 41 (2):123-28. hps://doi.org/10.1097/
OLQ.0000000000000075.
Barnabas,S.L.,Dabee,S.,Passmore,J-A.S.,Jaspan,H.B.,Lewis,D.A.,Jaumdally,S.
Z.,Gamieldien,H., etal.2017.ConvergingEpidemics ofSexuallyTransmied
InfeconsandBacterial VaginosisinSouthern AfricanFemaleAdolescentsat
RiskofHIV.Internaonal Journal of STD & AIDS.
Buvé,A.,Jespers,V.,Cruci,T.andFichorova,R. N.2014.TheVaginalMicrobiota
and Suscepbility to HIV. Aids 28 (16):2333-44. hps://doi.org/10.1097/
QAD.0000000000000432.
Callahan,B.J.,DiGiulio,D.B.,Goltsman,D.S.A.,Sun,C.L.,Costello,E.K.,Jeganathan,
P.,Biggio,J.R.,etal. 2017.ReplicaonandRenementofaVaginalMicrobial
Signature of Preterm Birth in TwoRacially Disnct Cohorts of US Women.
Proceedings of the Naonal Academy of Sciences114(37):201705899.hps://
doi.org/10.1073/pnas.1705899114.
Caporaso,J.G.,Kuczynski,J.,Stombaugh, J.,Binger,K.,Bushman,F.D.,Costello,
E. K., Fierer, N., et al. 2010. QIIME Allows Analysis of High-Throughput
CommunitySequencingData.Nature Methods7(5):335-36.
Edgar,R. C.2010. Searchand ClusteringOrdersof MagnitudeFasterthan BLAST.
Bioinformacs 26 (19):2460-61. hps://doi.org/10.1093/bioinformacs/
btq461.
Edgar,R. C.,Haas, B.J., Clemente,J. C.,Quince, C.andKnight,R. 2011.UCHIME
Improves Sensivity and Speed of Chimera Detecon. Bioinformacs 27
(16):2194-2200.hps://doi.org/10.1093/bioinformacs/btr381.
Feweis,J.M.,Brooks,J.P.,Serrano,M.G.,Sheth,N.U.,Girerd,P.H.,Edwards,D.
J.,Strauss, J.F,Jeerson, K.K., andBuck, G.A. 2014.Dierencesin Vaginal
Microbiome in African American Women versus Women of European
Ancestry. Microbiology (United Kingdom) 160 (2014):2272-82. hps://doi.
org/10.1099/mic.0.081034-0.
Feweis,J.M, Serrano,M.G., Sheth,N.U.,Mayer,C.M., Glascock,A.L., Brooks,
J.P.,Jeerson,K.K.and Buck,G.A..2012. Species-LevelClassicaonofthe
VaginalMicrobiome.BMC Genomics13(Suppl8):1-9.
Freitas,A. C., Bocking , A., Hill, J. E. and Money,D. M. 2018. Increased Richness
and Diversity of the Vaginal Microbiota and Spontaneous Preterm Birth.
Microbiome 6 (1). Microbiome:1-15. hps://doi.org/10.1186/s40168-018-
0502-8.
Gallo, M. F., Macaluso, M., Warner, L., Fleenor, M. E., Hook, E. W.,Brill, I. and
Weaver,M.A.2012.BacterialVaginosis,Gonorrhea,andChlamydialInfecon
AmongWomenAendingaSexuallyTransmiedDiseaseClinic:ALongitudinal
Analysis of Possible Causal Links. Annals of Epidemiology 22 (3):213-20.
hps://doi.org/10.1016/j.annepidem.2011.11.005.
Gaujoux,R.2014.GenerangHeatmapsforNonnegaveMatrixFactorizaon.
Gosmann,C., Anahtar, M.N.,Handley,S.A., Walker,B.D.,Virgin,H. W., Kwon,D.
S.,Gosmann, C.et al.2017. Lactobacillus-Decient CervicovaginalBacterial
CommuniesAre Associatedwith IncreasedHIV Acquisionin YoungSouth
African Women Report Lactobacillus -Decient Cervicovaginal Bacterial
CommuniesAreAssociatedwithIncreasedHIVAcquisionIn.Immunity46.
ElsevierInc.:1-9.hps://doi.org/10.1016/j.immuni.2016.12.013.
Holst, E., Goeng, A. R. and Andersch B. 1994. Bacterial Vaginosisand Vaginal
MicroorganismsinIdiopathicPrematureLaborandAssociaonwithPregnancy
Outcome.Journal of Clinical Microbiology32(1):176-86.
Jacobs,M.V.,Snijders,P.J.,Helmerhorst,T.J.,Meijer,J.andWalboomers,J.M.1997.
AGeneralPrimerGP5+/GP6+-MediatedPCR-EnzymeImmunoassayMethod
for Rapid Detecon of 14 High-Risk and 6 Low-RiskHuman Papillomavirus
GenotypsinCervicalScrapings.Journal of Clinical Microbiology35(3):791.
Kenyon, C.s, Colebunders, R. and Cruci, T. 2013. The Global Epidemiology of
Bacterial Vaginosis: A Systemac Review. American Journal of Obstetrics
and Gynecology 209 (6). Elsevier Inc:505-23. hps://doi.org/10.1016/j.
ajog.2013.05.006.
Leich,H. andKiss, H.2007. AsymptomacBacterial Vaginosisand Intermediate
Flora as Risk Factors for Adverse Pregnancy Outcome. Best Pracce and
Research: Clinical Obstetrics and Gynaecology 21 (3):375-90. hps://doi.
org/10.1016/j.bpobgyn.2006.12.005.
Lennard, K., Dabee, S., Barnabas, S. L., Havyarimana, E., Blakney A., Jaumdally,
S. Z., Botha, G, et al. 2017. Microbial Composion Predicts Genital Tract
Inammaonand PersistentBacterialVaginosis inAdolescentSouth African
Women. Infecon and Immunity, no. October:IAI.00410-17. hps://doi.
org/10.1128/IAI.00410-17.
Lewis,D.A, Müller, E.,Steele,L.,Sternberg, M.,Radebe,F.,Lyall,M.,Ballard, R.C.
and Paz-Bailey,G. 2012. Prevalence and Associaons of Genital Ulcer and
UrethralPathogensinMenPresenngwithGenitalUlcerSyndrometoPrimary
HealthCareClinicsinSouthAfrica.Sexually Transmied Diseases39(11):880-
85.hps://doi.org/10.1097/OLQ.0b013e318269cf90.
Liaw,A. andWiener,M. 2002.Classicaon andRegression byRandomForest.R
News2(3):18-22.hps://doi.org/10.1177/154405910408300516.
Mbulawa,Z. Z.A.,VanSchalkwyk, C.,Nai-chung Hu, Meiring,T.L., Barnabas,S.,
Dabee,S.,Jaspan,H.,etal.2018.HighHumanPapillomavirus(HPV)Prevalence
inSouth AfricanAdolescents andYoung WomenEncourages ExpandedHPV
VaccinaonCampaigns.PloS One13(1):1-15.
McClelland, R. S., Lingappa, J. R., Srinivasan, S., Kinuthia, J, John-Stewart, G.
C., Jaoko, W., Richardson, B. A., et al. 2018. Evaluaon of the Associaon
betweentheConcentraonsofKeyVaginalBacteriaandtheIncreasedRiskof
HIVAcquision inAfricanWomenfrom FiveCohorts:ANested Case-Control
Study.The Lancet Infecous Diseases3099(18).ElsevierLtd:1-11.hps://doi.
org/10.1016/S1473-3099(18)30058-6.
McMurdie, P. J. and Holmes, S. 2013. Phyloseq: An R Package for Reproducible
Interacve Analysis and Graphicsof Microbiome Census Data. PloS One 8
(4):e61217.hps://doi.org/10.1371/journal.pone.0061217.
Nelson,D. B.,Hanlon, A.L., Wu,G.,Liu, C.and DavidFredricks, D.N.2015. First
TrimesterLevels of BV-Associated Bacteria and Risk of Miscarriage Among
Women Early in Pregnancy. Maternal and Child Health Journal 19 (12).
SpringerUS:2682-87.hps://doi.org/10.1007/s10995-015-1790-2.
Oksanen,A.J.,Blanchet,F.G.,Kindt,R.,Legendre,P.,Minchin,P.R., Hara,R.B.O.,
Simpson,G.L., Solymos,P., Stevens,M.H.H., andWagner, H.2016.Package
‘Vegan.’
Paulson,J.N., Sne,O.C., Bravo,H.C. andPop,M. 2013.DierenalAbundance
Analysisfor Microbial Marker-Gene Surveys.Nature Methods 10 (12):1200-
1202.hps://doi.org/10.1038/nmeth.2658.
Pearce,M.M.,Hilt,E.E.andRosenfeld,A.B.2014.TheFemaleUrinaryMicrobiome :
AComparison ofWomenwith andwithoutUrgency UrinaryInconnence.5
(4):1-12.hps://doi.org/10.1128/mBio.01283-14.Editor.
Ravel,J.,Gajer,P.,Abdo,Z., Schneider, G.M.,Koenig, S.S.K., Mcculle,S.L., Ault,
K.,Peralta,L.andForney,L.J.2010.VaginalMicrobiomeofReproducve-Age
Women. PNAS 108:4680-87. hps://doi.org/10.1073/pnas.1002611107/-/
DCSupplemental.www.pnas.org/cgi/doi/10.1073/pnas.1002611107.
Sing, T.,Sander, O., Beerenwinkel, N. and Lengauer, T. 2005. ROCR: Visualizing
Classier Performance in R. Bioinformacs 21 (20):3940-41. hps://doi.
org/10.1093/bioinformacs/b623.

 hp://www.satnt.ac.za 9 OpenAccess
 Page9of9 OorspronklikeNavorsing
Srinivasan,S., Homan,N. G.,Morgan,M. T.,Matsen, F. A.,Fiedler,T.L., Hall,R.
W.,Ross, F.J., et al. 2012. Bacterial Communies in Womenwith Bacterial
Vaginosis : High Resoluon Phylogenec Analyses Reveal Relaonships of
MicrobiotatoClinicalCriteria.PloS One7(6).hps://doi.org/10.1371/journal.
pone.0037818.
Tabatabaei,N.,Eren,A.M.,Barreiro,L. B.,Yotova,V.,Dumaine,A., Allard,C.and
Fraser,W.D.2018.VaginalMicrobiomeinEarlyPregnancyandSubsequentRisk
ofSpontaneous PretermBirth: ACase-ControlStudy.BJOG: An Internaonal
Journal of Obstetrics & Gynaecology, 8-11. hps://doi.org/10.1111/1471-
0528.15299.
Vinturache,A.E.,Gyam-Bannerman,C.,Hwang,J.,Mysorekar,I.U.andJacobsson,
B. 2016. Maternal Microbiome – A Pathway to Preterm Birth. Seminars
in Fetal and Neonatal Medicine 21 (2):94-99. hps://doi.org/10.1016/j.
siny.2016.02.004.
Wiesenfeld,H.C., Hillier,S. L.,Krohn,M.A., Landers,D.V.andSweet, R.L.2003.
Bacterial Vaginosis Is a Strong Predictor of Neisseria Gonorrhoeae and
Chlamydia Trachomas Infecon. Clinical Infecous Diseases : An Ocial
Publicaon of the Infecous Diseases Society of America 36 (5):663-68.
hps://doi.org/10.1086/367658.
Zhou,X., Brown,C.J.,Abdo, Z.,Davis,C. C.,Hansmann,M. A.,Joyce,P.,Foster, J.
A.andForney,L.J.2007.DierencesintheComposion ofVaginalMicrobial
CommuniesFound inHealthyCaucasianand BlackWomen.ISME Journal1
(2):121-33.hps://doi.org/10.1038/ismej.2007.12.